../

app.py

@@ -28,22 +28,43 @@ darija_topic_labels = ["Customer Service", "Retention Service", "Billing Issue"]
 english_topic_labels = ["Support Request", "Subscription Issue", "Payment Dispute"]  # Adjust for English topics
 
 
+import torch
+
 def transcribe_audio(audio):
     """Convert audio to text, translate it, and classify topics in both Darija and English"""
-    audio_array, sr = librosa.load(audio, sr=16000)
-    input_values = processor(audio_array, return_tensors="pt", padding=True).input_values
+    try:
+        # Load and preprocess audio
+        audio_array, sr = librosa.load(audio, sr=16000)
+        
+        # Ensure correct sampling rate
+        input_values = processor(audio_array, sampling_rate=16000, return_tensors="pt", padding=True).input_values
+
+        # Move to GPU if available
+        device = "cuda" if torch.cuda.is_available() else "cpu"
+        model.to(device)
+        input_values = input_values.to(device)
+
+        # Get predictions from Wav2Vec2 model
+        with torch.no_grad():
+            logits = model(input_values).logits
+        tokens = torch.argmax(logits, axis=-1)
+
+        # Decode transcription (Darija)
+        transcription = processor.decode(tokens[0])
+
+        # Translate to English
+        translation = translate_text(transcription)
 
-    logits = model(input_values).logits
-    tokens = torch.argmax(logits, axis=-1)
+        # Classify topics for Darija and English
+        darija_topic = classify_topic(transcription, arabert_tokenizer, arabert_model, darija_topic_labels)
+        english_topic = classify_topic(translation, bert_tokenizer, bert_model, english_topic_labels)
 
-    transcription = processor.decode(tokens[0])
-    translation = translate_text(transcription)
+        return transcription, translation, darija_topic, english_topic
 
-    # Classify topics for both Darija and English
-    darija_topic = classify_topic(transcription, arabert_tokenizer, arabert_model, darija_topic_labels)
-    english_topic = classify_topic(translation, bert_tokenizer, bert_model, english_topic_labels)
+    except Exception as e:
+        print(f"Error in transcription: {e}")
+        return "Error processing audio", "", "", ""
 
-    return transcription, translation, darija_topic, english_topic
 
 
 def translate_text(text):